← Back02 · AI visibility
Built for the agents that now do the searching. If machines can’t read you, you’re invisible at the moment of decision.
01 · Why it exists
The SEO keyword
You optimized to place a link a human would scan, click, and read. The page was the destination.
AI visibility
Now an agent reads the web and answers in one voice. It names a few brands. If it can’t read you, you’re not in the room where the decision happens.
Your customers ask AI before they ask you. The sales conversation is already happening. The only question is whether you’re in it.
2026 market dataHousehold names surface in 73% of first-run AI answers. Niche and regional brands: 11%.
03 · The case
Cetrogar. A 46-year-old retailer: 92 stores, 15 provinces. Here is exactly how the agents saw it.
Your brand is invisible to the agents. Today, AI’s answers are built without you.
Home renders 11 words without JavaScript. Category pages: zero words, no title. Only product pages are legible.
46 years, 92 stores, 15 provinces, and none of it is extractable. Zero definitions an agent can quote.
Engines know what it is, but a tax registry and third parties define it. No sameAs, no owned source.
Complaint sites rank in the top 10 sources for "is it trustworthy". The narrative is written by others.
Fresh catalog, zero editorial. No authors, no dates, no Person schema.
In the category questions, where a new customer is captured, Frávega appears in 49% of the answers. Cetrogar, in 9%. Five times less.
When asked by name, Cetrogar shows up 100% of the time, in all 19 brand runs. It is not that it doesn’t exist. It doesn’t appear where the purchase is decided, and when it does, complaint sites write the narrative.
“What is the best appliance store in Argentina?” · asked to Perplexity
It recommends Frávega (“a very widespread chain, with deals and installments”) and Naldo by name. The second source cited is fravega.com, whose legible site cites itself.
Cetrogar does not appear in this answer.
Run 1 of 5 · 2026-06-11 · consistent across runs
“Is Cetrogar trustworthy to buy online?” · asked to Perplexity
It always answers, but the answer is “opinions are very divided, a significant volume of complaints, delays, dented fridges,” citing tuquejasuma and defensadelconsumidor.
It wins the mention and loses the narrative.
Run 1 of 5 · 2026-06-11 · complaint sites as primary source
Scope, stated plainly: this sample ran 1 of the 4 engines (Perplexity), 24 prompts at 3 to 5 runs each, for an API cost near US$1.72. The full audit runs ChatGPT, Claude, Gemini and Perplexity with a formal LLM judge. Site findings hold regardless of engine.
Viñas de Balbo. The agents can half-read it. Outside its own name, across 45 category runs, it appeared in exactly one. Santa Julia and Dada, which don’t even compete in its format, own the category answer. A different brand, the same silence.
04 · How it works
We crawl your key pages the way GPTBot and ClaudeBot do. If the content only exists after JavaScript, the agents never see it.
We run the real questions your customers ask, across ChatGPT, Claude, Gemini and Perplexity, several times each, with dates. No cherry-picking.
One number, 0 to 100, across five weighted dimensions: access, citable content, entity, third-party surface, freshness. Frozen and defensible.
The shortlist that makes you readable, citable and cited: server-rendered facts, definitions, schema, and ownership of the category answer.
The searching moved to the machines. Be the answer they give, not the brand they forget.